Postal logistics organizations are characterized as having high labor intensity and short response times. These characteristics, along with rapid change in mail volume traffic, make load scheduling a fundamental concern. Load analysis of major postal infrastructures such as post offices, sorting centers, exchange centers, and delivery stations is required for optimal postal logistics operation. Especially, the performance of postal traffic forecasting is essential for optimizing the resource operation by accurate load analysis. Therefore, this paper addresses a demand forecasting problem for parcel logistics. The main purpose of this paper is to describe a machine learning approach for predicting short-term traffic of postal parcel based on feature engineering and to introduce an application to on-site logistics service of Korea Post. The proposed method consists of three main phases. First, the characteristics of the postal traffic are analyzed and calendar and volume-based features are generated. Second, multiple regression models by the clusters resulted from feature engineering are developed. Finally, individual models for level 4 and level 5 delivery stations are constructed to reinforce prediction accuracy. The experiment shows the advantage in terms of forecasting performance. Comparing with other techniques, experimental results show that the proposed scheme improves the average performance up to 50.1%.
